Why did the QV value decrease significantly after using YAHS for HiC scaffolding?

[JieQYan]

Hi,I input the genome purdged.fa (QV=52) assembled by hifiasm and purged by haplotig into YAHS, and the resulting scaffolds_final.fa was directly reduced to 23 after passing through Merqury,I want to know what caused this?
Below is my code：

#!/bin/bash
set -e

# --- Set variables (please check paths and filenames) ---
ASSEMBLY_FA="purged.fasta"
HIC_R1=/users/jieqyan/workspace/assembly/ref/Hi-C/AT_HiC_R1.clean.fastq.gz
HIC_R2=/users/jieqyan/workspace/assembly/ref/Hi-C/AT_HiC_R2.clean.fastq.gz
# Note: PICARD_JAR variable removed, using 'picard' as the executable command
THREADS=20

echo "--- Stage 1: Hi-C preprocessing begins ---"

# 1. Index the cleaned contigs
echo "1. Indexing the cleaned contigs..."
bwa index $ASSEMBLY_FA
samtools faidx $ASSEMBLY_FA

# 2. Align Hi-C reads and generate sorted BAM
echo "2. Aligning Hi-C reads and generating sorted BAM..."
bwa mem -5SP -t $THREADS $ASSEMBLY_FA $HIC_R1 $HIC_R2 | \
samtools view -@ 4 -bS - | \
samtools sort -@ 4 -o hic.sorted.bam

# 3. Mark and remove PCR duplicates
echo "3. Picard MarkDuplicates removing duplicates..."
# Ensure -Xmx comes before, using 'picard' command
picard -Xmx32g MarkDuplicates \
    I=hic.sorted.bam \
    O=hic_dedup.bam \
    M=metrics.txt \
    REMOVE_DUPLICATES=true

# 4. Filter and sort by read name
echo "4. Filtering non-unique mappings and sorting by name..."
samtools view -F 0x904 -h -b hic_dedup.bam | \
samtools sort -@ 4 -n -o hic_clean.namesort.bam

echo "--- Stage 1 completed. Output: hic_clean.namesort.bam ---"

yahs purged.fasta hic_clean.namesort.bam -o yahs_scaffolding_output
I would be very grateful if you could offer me any advice！

[c-zhou]

Hello @xinguok,

Do you have stats for purged.fasta and scaffolds_final.fa such as number of bases, number of sequences, N50?

YaHS does not change the original contig sequence content - only breaks contigs and add gaps between contigs to construct scaffolds, so the QV should not change very much. Are you sure the YaHS and Merqury runs were successful?

Best,
Chenxi

[JieQYan]

Thank you very much for your prompt reply！
In addition to QV, I also evaluated LAI and busco. The results showed that both metrics for scaffolds_final.fa were quite good (LAI=25, busco=100), but QV remained low. Strangely, after polishing scaffolds_final.fa again, the QV value returned to 52, but scaffolds_final.fa became more fragmented. It seems that the significant decrease in QV value after HiC scaffolding is an illusion. Therefore, I would like to ask if YAHS performed some kind of special masking on the genome, making it impossible for Merqury to correctly identify bases during genome evaluation, thus causing a significant decrease in QV value?

[foreignsand]

To the OP, @xinguok, did you ever figure out what was happening with this? I'm having the same problem and am also wondering if it's something to do with merqury not behaving well with scaffolded assemblies.

[foreignsand]

Hi again, @xinguok! I wanted to let you know that for me, the problem has been resolved. You aren't going to believe this, but if it's the same issue I had, then it will just be that the yahs FASTA wraps lines while hifiasm does not. Try linearizing the yahs FASTA and rerunning merqury.

See here:
#122

Cheers,
Emily